DeepYardDeepYard
V

Vera

Automated safety testing framework for LLM agents with risk discovery and verification

Open SourceFree

About

End-to-end framework for systematically testing LLM agent safety at scale. Features a three-stage pipeline: automated risk discovery to identify potential failure modes, test case generation targeting those risks, and evidence-grounded verification of agent behavior. Designed for researchers and developers building production AI agents who need comprehensive safety evaluation beyond basic benchmarks.

Details

Type
Integrations
Language

Tags

evaluationautonomousframeworkopen-sourceobservability